There is consensus that plant species richness enhances plant productivity within natural grasslands, but the underlying drivers remain debated. Recently, differential accumulation of soil-borne fungal pathogens across the plant diversity gradient has been proposed as a cause of this pattern. However, the below-ground environment has generally been treated as a 'black box' in biodiversity experiments, leaving these fungi unidentified. Using next generation sequencing and pathogenicity assays, we analysed the community composition of root-associated fungi from a biodiversity experiment to examine if evidence exists for host specificity and negative density dependence in the interplay between soil-borne fungi, plant diversity and productivity. Plant species were colonised by distinct (pathogenic) fungal communities and isolated fungal species showed negative, species-specific effects on plant growth. Moreover, 57% of the pathogenic fungal operational taxonomic units (OTUs) recorded in plant monocultures were not detected in eight plant species plots, suggesting a loss of pathogenic OTUs with plant diversity. Our work provides strong evidence for host specificity and negative density-dependent effects of root-associated fungi on plant species in grasslands. Our work substantiates the hypothesis that fungal root pathogens are an important driver of biodiversity-ecosystem functioning relationships.